False-negative and False-positive Diffusion-weighted MR Findings in Acute Ischemic Stroke and Stroke-like Episodes

JRural Med 2005 ; 1 : 27 32 Original Article False-negative and False-positive Diffusion-weighted MR Findings in Acute Ischemic Stroke and Stroke-like Episodes Shuzo Shintani 1,HiroakiYokote 1,KaoruHanabusa 2 and Tatsuo Shiigai 3 Departments of 1 Neurology, 2 Diagnostic Radiology, and 3 Internal Medicine, Toride Kyodo General Hospital, Ibaraki, Japan Abstract Background : Diffusion-weighted magnetic resonance (MR) imaging (DWI) is an excellent examination for detecting acute ischemic stroke, but false-negative cases have been reported recently. Patients and Methods : Since the present MR scanner (1.5-T, Siemens Symphony) was introduced to our hospital, a prospective study was designed in the Departments of Neurology and Radiology to evaluate the DWI findings in patients tentatively diagnosed to have an acute infarction and in those with stroke-like episodes. During the 31 months between June 2000 and December 2002, 572 consecutive patients with acute cerebral infarction or presenting conditions mimicking ischemic stroke, including transient ischemic attack (TIA), sudden-onset, and loss of consciousness (LOC) with or without seizure, underwent DWI. Results : Four of 366 patients with a cerebral infarction (1.1%) had false-negative DWI in the acute stage, and 10 of 206 patients with conditions mimicking ischemic stroke (4.9%) had false-positive DWI in the acute stage. Of these 10 patients, there were five cases with TIA, four with sudden-onset, and 1 with LOC with seizure. Sensitivity and specificity values were 98.9% and 97.6%, respectively, when DWIs were performed to diagnose acute cerebral infarction. Conclusion : DWI rarely fails to detect an acute-stage cerebral infarction, but further confirmatory measures may be necessary when there is a negative examination using a clinical or computed tomographic diagnosis to the contrary. Key words : acute ischemic stroke, stroke-like episode, magnetic resonance imaging, diffusionweighted imaging, ischemic penumbra Correspondence to : Shuzo Shintani Department of Neurology, Toride Kyodo General Hospital, 2-1-1 Hongoh, Toride City, Ibaraki 302-0022, Japan E-mail : dw4s-sntn@asahi-net.or.jp Introduction Diffusion-weighted magnetic resonance ( MR ) imaging (DWI) is much more sensitive than conventional T2-weighted MR imaging or computed tomography ( CT) for detecting early changes associated with hyperacute ischemic stroke ; the reported sensitivity approaches 100%. Recently, however, several false-negative cases have been reported 1~3).In1999, Ay et al. identified 27 cases among 782 consecutive patients who presented with strokelike neurologic deficits but DWI in the area relevant to the symptoms was found to be normal 1).Inthesame year, Wang et al. reported on two patients with large acute cortical infarcts but negative DWI and discussed possible mechanisms 2). In 2000, Oppenheim et al. described false-negative DWI in eight of their acute stroke cases (5.8%), including five with brain-stem infarcts and three with cerebral lesions 3). We report on the prevalence of false-negative acutestage DWI studies in patients with persistent neurologic deficits due to ischemic stroke, and the putative causes why DWI initially missed the lesion will be discussed. Methods and Patients Methods The present MR scanner (1.5-T, Siemens Symphony) has been in use in our hospital since June 2000. Multislice whole-brain DWI was performed in patients provisionally diagnosed to have acute ischemic stroke or conditions mimicking ischemic stroke. DWI were obtained in the axial plane by combining a single-shot spin-echo (SE)-type echo planar imaging (EPI) sequence and additional motion probing gradient (MPG) pulses in three orthogonal axes. Imaging parameters were as follows : echo time (TE) 154 msec ; matrix, 128 128 ; field of view (FOV), 23 23cm ; slice thickness, 5mm ; and intersection gap, 1.5mm. The b- values were 0 and 1000 sec/mm 2. Other routine MRI sequences included the following : axial spin-echo (SE) T1-weighted images 2005 The Japanese Association of Rural Medicine 27

(TR/TE=650/16, matrix 256 192, FOV=23 23cm, slice thickness=5mm, 1.5-mm gap, 1 signal acquisition) ; axial Turbo spin echo (TSE) T2-weighted images (TR/ TE=4500/98, matrix 256 192, FOV=23 23cm, slice thickness=5mm, 1.5-mm gap, 1 signal acquisition). Each imaging series consisted of 20 sequential sections of the entire brain. Patients At our hospital, it is routine to perform brain CT on admission to the emergency room (ER) as a first-line diagnostic evaluation for patients with sudden-onset neurologic deficits. These patients have, by definition, positive scores on the National Institute of Health Stroke Scale (NIHSS) 4) on admission. Those without neurologic deficits on admission have had positive NIHSS scores when they first developed the attacks prior to the present admission. When CT reveals no abnormalities, MR studies are recommended by an ER physician. DWI is another auxiliary choice to reliably detect ischemic lesions in an extremely acute stage. Following the introduction of Siemens Symphony, a prospective study was designed in the Departments of Neurology and Radiology to evaluate the DWI findings in patients provisionally diagnosed to have acute cerebral infarction or conditions mimicking ischemic stroke since June 2000. This study was approved by the institutional review committee of the hospital and was performed between June 2000 and December 2002. The MR findings were evaluated twice, by both a registered radiologist (KH) and a neurologist (SS). During the 31 months of the study, DWIs were performed on 572 consecutive patients with acute cerebral infarction or with conditions mimicking ischemic stroke. The former group of patients were confirmed to have positive neurologic deficits and the latter to not have neurologic deficits on admission. The latter group of patients consisted of those with transient ischemic attack (TIA), sudden-onset isolated vertigo, or loss of consciousness (LOC) with or without seizure. Considering the diagnostic precision inherent in DWI to detect lesions of acute cerebral infarction, falsepositive and false-negative cases were defined as follows : the former presented with neurologic deficits but without DWI findings and the latter pressented without neurologic deficits but with DWI findings on admission. Results Four of 366 patients with cerebral infarction (1.1%) had false-negative DWI in the acute stage (Table 1 ; Figures 1 and 2). Ten of 206 patients with conditions mimicking ischemic stroke (4.9%) had false-positive DWI in the acute stage (Table 2 ; Figure 3). Of these patients, there were five with TIA, four with suddenonset, and one with LOC with seizure. Sensitivity and specificity values for detecting lesions of acute cerebral infarction by DWI were 98.9% and 97.6%, respectively (Table 3). Case Reports Patient 1 (false-negative case, Figure 1) A 69-year-old man with a history of hypertension and cerebral infarction 3 years prior to admission developed right hemiparesis. He had been prescribed ticlopidine 200mg daily to decrease the risk of a recurrent infarct as well as an antihypertensive drug (Table 1). On admission, his blood pressure was 166/ 101 mmhg, and his heart rate was 85/min with a regular rhythm. The patient was alert, and his cranial nerves were intact. He presented with right hemiparesis, together with a slight left hemiparesis remaining. Cerebellar function and deep tendon reflexes were bilaterally normal, but a positive Babinski sign was present on the right side. CT and DWI taken 10 h after the onset of symptoms showed no abnormalities except for the previous infarct, located in the right corona radiate (Figure 1). MR angiography (MRA) revealed no abnormalities in the major vessels of the brain and neck. The right hemiparesis persisted after admission. Follow-up DWI 9 days after the onset detected a highintense signal lesion in the left corona radiata (Figure 1). Patient 4 (false-negative case, Figure 2) A 94-year-old woman with a history of hypertension and transient ischemic attack (TIA) 1 year previously suddenly developed congnitive dysfunction and developed paresis of left hand. She had been prescribed ticlopidine (200mg) and aspirin (100mg) daily to reduce risk of cerebral vascular events, and also an antihypertensive drug (Table 1). On admission the blood pressure was 178/74 mmhg, and the heart rate was 88/min with a regular rhythm. She exhibited paresis of left hand, and the consciousness was not clear. She was unable to respond correctly to questions. CT and DWI performed 22 h after onset of symptoms showed no abnormalities. MRA revealed no abnormalities in major vessels of the brain or neck. After admission she showed persistent paresis of the left hand. Follow-up DWI 2 days after onset of symptoms detected a lesion with signal hyperintensity involving the right temporal, parietal, and occipital lobes (Figure 2). Patient 5 (false-positive case, Figure 3) This 70-year-old male with TIA revealed no 28

Table 1 Findings of 4 patients with sudden neurologic deficit and negative initial findings on diffusion-weighted MR study Case No. Age (y) Gender Stroke risk factor New neurologic symptoms 1 69/M Hypertension Sudden R hemiparesis 2 77/F Hypertension Atrial fibrillation 3 53/M Hypertension Diabetes mellitus Sudden L hemiparesis Sudden R hemiplegia 4 94/F Hypertension Sudden L upper limb paresis and acute onest dementia History of previous cerebral infarction (+) 3yearsago (+) 8monthsago (+) 5yearsago TIA 1yearago Residual neurologic deficits Lhemiparesis Amnesic aphasia Lhemiparesis None Prescribed durg with onest of new symptom Ticlopidine Warfarin Insulin Aspirin Ticlopidine Initial MR Findings MR Latency DWI (hr) MR Latency (days) Follow-up MR Findings DWI 10 Negative 9 Hyperintense signal of L corona radiata 3.5 Negative 2 Hyperintense signal of cortex and white matter of R parietal lobe 1 Negative 2 Hyperintense signal of L medulla 22 Negative 2 Hyperintense signal of cortex and white matter of R temporoparietooccipital lobe R:right;L:left ; TIA : transient ischemic attack ; MR : magnetic resonance ; DWI : diffusion-weighted image ; hr : hour (s). Figure 1 (Patient 1, false-negative case) In a 69-year-old man with sudden onset of right hemiparesis, diffusion-weighted imaging (DWI) was negative 10 h after onset of symptoms (left) ; 9 days later (right), a lesion in the left corona radiata (arrowhead) shows a hyperintense signal. Figure 2 (Patient 4, false-negative case) In a 94-year-old woman with sudden onset of left upper limb paresis and acute cognitive deterioration, diffusion-weighted imaging (DWI) was negative 22 h after onset of symptoms (left) ; 2 days later (right), hyperintense signals were seen in the cortex and white matter of portions of the right temporal, parietal, and occipital lobes (arrowheads). 29

Table 2 Findings of 10 patients with negative neurologic deficits and positive initial findings on diffusionweighted MR study Case No. Diagnosis 5 TIA (1-hr duration) Age (y) Gender New neurologic deficits on admission Transient symptoms 70/M Negative Left arm paresis Stroke risk factor Residual neurologic deficits Prescribed drug with onset of transient symptom MR Latency (hr) Intensity in ADC map Negative Negative None 2 low 6 TIA (3-hr duration) 7 TIA (5-hr duration) 8 TIA (30-min duration) 9 TIA (10-min duration) 10 Sudden onset 11 Sudden onset 12 Sudden onset 13 Sudden onset 14 LOC with seizure 74/M Negative Left leg paresis Hypertension Negative 80/F Negative Numbness of right arm and dysarthria 68/M Negative Right arm paresis 71/M Negative Right leg paresis Hypertension, old cerebral infarct Hypertension, Atrial fibrillation Dementia Negative Ticlopidine, Aspirin, Hypertension Negative Diabetes Negative Anti-diabetic mellitus 48 low 185 equal 3 equal 15 equal 15 (artifact) 170 low 80/F Negative Vertigo Negative Negative None 16 equal 59/M Negative Vertigo Hypertension, Diabetes mellitus 64/M Negative LOC and convulsion Negative 58/M Negative Vertigo and nystagmus 74/M Negative Vertigo Hypertension Negative, Insulin 14 low Negative Negative None 10 low TIA : transient ischemic attack ; LOC : loss of consciousness ; ADC : apparent diffusion coefficient ; hr : hour (s) ; min : minutes. neurologic deficits on admission (Table 2), but DWI showed a positive high-intensity region. Figure 3 (Patient 5, false-positive case) Diffusion-weighted imaging was positive in the acute stage in a patient with a transient ischemic attack (TIA). Discussion Since 1999, there have been reports of several patients with acute cerbral infarction who had falsenegative DWI 1~3). Ay et al. identified 27 cases among 782 consecutive patients 1).Theyhadneurologic deficits on admission, but DWIs revealed no corresponding abnormalities. However, the final diagnosis was conditions mimicking stroke in 10 of the patients (migraine, seizures, transient global amnesia, or brain tumor) and a cerebral ischemic event in only 17 patients (2.2%). The latter included 7 patients with lacunar syndrome (including three with infarcts demonstrated subsequently) and 10 patients with hemispheric cortical syndromes (five with TIA, two with prolonged reversible deficits, and three with 30

Table 3 Accuracy of DWI findings of MRI in an acute stage of cerebral infarction and in a condition mimicking ischemic stroke Acute stage of cerebral infarction Positive neurological deficits when MRI was Performed on admission Transient ischemic attack (TIA) Sudden onset Loss of conseiousness with or without seizure Negative neurological deficits when MRI was performed on admission Patients with positive DWI findings in initial MRI 362 5 4 1 Patients with negative DWI findings in initial MRI 4 17 91 88 DWI : diffusion-weighted image ; False-negative cases ; False-positive cases ; Sensitivity=98.9% ; Specificity=97.6%. infarction on follow-up imaging). In that year and the following year, Wang et al. and Oppenheim et al. described false-negative DWIs in acute ischemic stroke cases and discussed possible mechanisms 2, 3). They concluded that vertebrobasilar stroke cannot be ruled out by a negative early DWI, particularly when symptoms suggestive of this diagnosis persist. In our four patients (Patients 1 to 4), follow-up T2- weighted images showed high intensity regions corresponding to the hyper-intense lesions in the follow-up DWIs. These positive findings in follow-up T 2-weighted images in areas accounting for the neurologic defictis suggest that the patients had ischemic events. Lesion sites varied in our four patients. We concluded that the occluded vessels were a perforating artery in the white matter of the corona radiata in patient 1 (thrombosis), a parietal branch of the right middle cerebral artery (MCA) in patient 2 with AF (embolism), a left-sided perforating artery of the medulla in patient 3 ( thrombosis ), and a temporoparietal branch of the right MCA in patient 4 (possible embolism). The prevalence of false-negative DWI among acute infarcts at our hospital was 1.1% (4 in 366), which is somewhat lower than previous research reporting a 2.2% 1) or 5.8% 3) prevalence. In the current study we addressed not only the falsenegative rate but also the false-positive rate. The values of sensitivity and specificity were 98.9% and 97.6%, respectively, when DWIs were performed for the diagnosis of acute cerebral infarction (Table 3). First, five patients (Patients 5 to 9) with TIA had small hyperintense signal regions in the initial DWI on admission (Table 2). Transient symptoms were ipsilateral hemiparesis with and without dysarthria and numbness of the arm. MR latencies between the onset of symptoms and MR examination varied from 2 to 185 hours. The apparent diffusion coefficient (ADC) maps showed hypo-intense signal regions in Patients 5 and 6, but iso-intense signal regions in Patients 7, 8, and 9 in the corresponding DWI positive regions. Etiologies of these lesions have been discussed TIArelated DWI abnormalities are associated with prolonged TIA duration and disturbance of higher brain function 5, 6). The durations of TIA symptoms in our patients varied from 10 min to 5 hr. The relatively prolonged duration (1 hr to 3 hr) in Patients 5 and 6 might have resulted in the hypo-intense signals, and the relatively short duration (10 min to 30 min) in Patients 8 and 9 might have produced the iso-intense signals in the ADC maps. However, in Patient 7 the markedly prolonged duration (5 hr) resulted in the isointense signals and the failure to present hypo-intense signal regions on the ADC map. Second, four patients (Patients 10 to 13) with sudden onset had tiny hyperintense signal regions in the initial DWI (Table 2). They had no ataxia, paresis, or other symptoms of brain-stem involvement on admission. MR latencies varied from 14 to 170 hours. The lesions were located in the cerebellar peduncle, the vermis, and the corpus callosum. The ADC maps showed hypo-intense signals in Patients 11 and 13 but an iso-intense signal in Patient 12. These results were not related to the intensities and durations of their symptoms. In Patient 10, details were not evaluated by the artifacts. Those patients might have experienced ischemic stroke in the infratentorial area. Subsequent reperfusion could have resulted in these small lesions on admission. Third, the patient (Patient 14) with LOC with seizure had a small hyperintense signal region in the temporal lobe on the initial DWI (Table 2). The ADC map revealed a hypo-intense signal in the corresponding region. He was conscious and had no neurological deficits on admission. Lansberg et al. reported that some of the MRI and CT findings during patrtial status epilepticus mimicked those of acute ischemic stroke 7). Wang et al. suggested three possible mechanisms to explain the lack of diffusion changes in the acute phase in patients eventually proven to have infarction 2).First, cerebral blood flow (CBF) may have been at an intermediate level, below the threshold for neuronal 31

dysfunction (symptom onest) but above the threshold for reduced diffusion. Second, reperfusion may have occurred, restoring the diffusion constant to normal but not preventing eventual delayed infarction. Third, asecondischemicevent may have caused the eventual infarction. To varying extents, the above three hypotheses all involve the ischemic penumbra concept 8). When an ischemic stroke occurs, the peripheral vessels of the lesion initially dilate to sustain blood filling. Such compensatory vasodilation occurs in regions with mild reductions in CBF, leading to an increase in collateral blood flow and maintenance of normal cerebral blood volume (CBV). However, the reduced level of CBF may not be sufficient to maintain normal electrical activity and is reflected by functional neurologic deficits. When this autoregulatory mechanism is unsuccessful because of an excessive decrease in the blood filling pressure so that CBF becomes notably and remarkably decreased in the involved area, the oxygen extraction fraction (OEF) must be increased to maintain the oxygen supply inthisregion 9). This phenomenon is sometimes referred to as misery perfusion. It occurs in the penumbra and dose not represent irreversible ischemic involvement but rather a form of vasogenic edema of the brain. However, when the OEF reaches its upper limit, the cerebral metabolic rate for oxygen (CMRO2) begins to decrease, and subsequently, irreversible ischemic change with cytotoxic edema produces a DWI-positive area by MR 10). We suggest that the prolonged penumbra state due to unknown reasons resulted in the false-negative DWIs in this study. Reference 1) Ay H, Buonanno FS, Rordorf G, et al. Normal diffusion-weighted MRI during stroke-like deficits. Neurology 1999 ; 52 : 1784 1792. 2) Wang PY, Barker PB, Wityk RJ, et al. Diffusionnegative stroke : A report of two cases. AJNR Am J Neuroradiol 1999 ; 20 : 1876 1880. 3) Oppenheim C, Stanescu R, Dormont D, et al. Falsenegative diffusion-weighted MR findings in acute ischemic stroke. AJNR Am J Neuroradiol 2000 ; 21 : 1434 1440. 4) Lyden P, Brott T, Tilley B, et al. Improved reliability of the NIH Stroke Scale Using Video Training. Stroke 1994 ; 25 : 2220 2226. 5) Rovira A, Rovira-Gols A, Pedraza S, et al. Diffusion -weighted MR imaging in the acute phase of transient ischemic attacks. AJNR Am J Neuroradiol 2002 ; 23 : 77 83. 6) Inatomi Y, Kimura K, Yonehara T, et al. DWI abnormalities and clinical characteristics in TIA patients. Neurology 2004 ; 62 : 376 380. 7) Lansberg MG, O Brien MW, Norbash AM, et al. MRI abnormalities associated with partial status epilepticus. Neurology 1999 ; 52 : 1021 1027. 8) Hossmann K. Viability thresholds and the penumbra of focal ischemia. Ann Neurol 1994 ; 36 : 557 565. 9) Powers WJ. Cerebral hemodynamics in ischemic cerebrovascular disease. Ann Neurol 1991 ; 29 : 231 240. 10) Kelly PJ, Hedley-Whyte ET, Primavera J, et al. Diffusion MRI in ischemic stroke compared to pathologically verified infarction. Neurology 2001 ; 56 : 914 920. 32